Topical authority is built in clusters.

The starting point: real queries, never a hunch about a subject

This is where most store blogs fail: the subject gets chosen because it "seems relevant". A cluster built that way ranks on questions nobody asks. So the search data for your market is interrogated first, and failing that the queries the store already ranks on — and if there is no material at all, we say so instead of inventing an angle.

• Market queries with their volume and their difficulty, before any subject is arbitrated.
• A documented fallback onto the queries the store already captures when the market returns nothing.
• Subjects too close to an article already published are excluded: two articles targeting the same query cannibalise each other.

One pillar, four satellites, and linking that does not read as mechanical

Every cluster runs to four or five articles, the first of which is the pillar covering the broad subject. Each satellite addresses one precise search intent. The pillar is written and published first: a cluster whose pillar is not live makes no sense, and if its publication fails, no satellite is attempted.

• The pillar first, always — it is what carries the cluster's authority.
• Sections of roughly 500 words, outline then introduction then body, one article at a time.
• Internal linking is drawn section by section: each link is only placed one time in two. That variation is what makes the linking read as natural rather than as a template repeated identically — linking that is too regular is easy to spot.
• Every cluster points at the pages that convert: a cluster leading to no product page is decorative.

Nothing is written to the store if the quality check blocks

Before an article touches your site, a quality check verifies what conditions its ranking: heading structure, length and uniqueness of the title tag and meta description, density, internal linking, presence of the expertise signals. If it blocks, the article is not created — it is corrected and put through again. An override exists but is only used on an explicit human decision, never as routine.

• A blocking check: a non-compliant article never reaches the store.
• Structured markup generated systematically — article, identifiable author, frequently asked questions, breadcrumb. Never hand-written: one badly escaped quotation mark and the whole block is ignored by search engines, without a single visible error.
• Questions declared in the markup must exist in the page with their answer. Declaring an FAQ that is not there is a false declaration, and is penalised as one.

Everything goes out as a draft, nothing goes live without your approval

The cluster is created as a draft on your store, then submitted for approval on your dedicated channel. You read it, you ask for a rewrite or you give the go — and only then does publication proceed. A cluster awaiting approval blocks the next one from opening: with two conversations in flight on the same channel, nobody knows any more what they are answering.

• Drafts created on your store, visible before any publication.
• Explicit approval per cluster, with the option to have it rewritten.
• A rejected cluster does not consume the period's quota.
• Publishing straight to live is possible, but it has to be asked for: it is never the default setting.

Product pages matter just as much as the blog

The blog attracts, product pages convert — and they rank too. New ones are produced in the format that captures search, existing ones reworked starting with the highest-margin products: a title matching how people actually search, a description that answers objections, structured attributes, light visuals.

• 2 to 8 new optimised product pages per month depending on tier.
• Gradual rework of the existing catalogue, prioritised by margin and not by age.
• Full product markup: price, availability, reviews, delivery time.
• Duplicate content across variants dealt with — it quietly penalises a lot of catalogues.

What we refuse to produce

We do not publish for the sake of publishing. One more cluster on a subject unrelated to what you sell earns nothing, however much traffic it makes. If a month is better spent reworking twenty product pages than shipping another cluster, we propose that instead, and accept that the article count is lower that month.